gornishanov/await

await's Issues

Could I add the demo of CppCon2018 to the llvm test suite ?

Hi Gor,

Now the llvm lacks a benchmark to measure the performance for coroutines. I think this one is relatively suit for that.
And I am worrying about the license problem. Do you feel comfortable if I submit the code to llvm test suite (I would mark you as the author).

CppCon2018: nanotest

Dear Gor,
as VS16.8 has now official coroutine support i compiled your source on Windows and VS 2019 16.8.2 and Win10 x64. I just removed the experimental from the namespace,

All 3 algorithms of nanotest.cpp report "!!!! BUG" on windows and i cant understand why ....

naive big 1

count: 67108864 lookups: 1048576 repeat 5
50.106 ns per lookup/log2(size)
sum 2618390
!!!! BUG, expected 2624940

---

sm big 1

count: 67108864 lookups: 1048576 repeat 5
64.5962 ns per lookup/log2(size)
sum 2618390
!!!! BUG, expected 2624940

---

coro big 1

count: 67108864 lookups: 1048576 repeat 5
64.6175 ns per lookup/log2(size)
sum 2618390
!!!! BUG, expected 2624940

A question in 2016_CppCon/Coroutines Under The Covers - Gor Nishanov - CppCon 2016.pdf

The argument n should be copied and stored in the f.frame?

When resumed, n is still visted in function f, so it should be store in the f.frame. In another word, every stack based context storage(including the function arguments, local variables, for abi in x86) in f should be allocated in f.frame or copyed to f.frame.

webassembly

@GorNishanov are there any webassembly examples which exercise co_await?